HAL Id: hal-01113181
https://hal.archives-ouvertes.fr/hal-01113181
Preprint submitted on 4 Feb 2015
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
A complete methodology for determining the influence of the design for 3.3 kV silicon carbide diodes : JBS
compared to Schottky
Florian Chevalier, Pierre Brosselard, Dominique Planson, Pascal Bevilacqua, Grégory Grosset, Lionel Dupuy
To cite this version:
Florian Chevalier, Pierre Brosselard, Dominique Planson, Pascal Bevilacqua, Grégory Grosset, et al..
A complete methodology for determining the influence of the design for 3.3 kV silicon carbide diodes : JBS compared to Schottky. 2013. �hal-01113181�
A complete methodology for determining the influence of the design for 3.3 kV silicon carbide diodes : JBS compared to
Schottky
Florian Chevaliera,c, Pierre Brosselarda, Dominique Plansona, Pascal Bevilacquaa, Gr´egory Grossetb, Lionel Dupuyb
aUniversit´e de Lyon - Laboratoire Amp`ere INSA de Lyon
21 av. Jean Capelle
69 621 Villeurbanne CEDEX - France
bIon Beam Services avenue Gaston Imbert prolong´ee
13 790 Peynier - France
cflorian.chevalier@insa-lyon.fr
Abstract
The paper presents the design and the characterization of Schottky and JBS diodes, for high voltage matching. Design for manufacturing has been preferred and diodes have been fabricated following a semi-industrial methodology. Then electrical characterization was performed on devices : static characterization, including temperature influence, dynamic characterization in a buck-like con- figuration to evaluate the behavior in switching mode and C-V characterization to identify some parameters like the junction area. Whereas forward conduction is provided with low on-state re- sistance, 3.3 kV capability is demonstrated with low leakage current, low turn-on and low turn-off delays.
Keywords: Buck converter, Edge termination, High integration level, High voltage, JBS diodes, Power devices, Schottky diodes, Silicon carbide.
Introduction
Today’s trend is to save energy and a widely used energy carrier is electricity. Great projects such as smart grid networks need static power converters with high voltage capa- bilities. Devices with high performances and low loss are needed. In this context, this pa- per presents the design, the fabrication and the characterization of 3.3 kV high voltage 4H- SiC Schottky and JBS diodes based on a sim-
ple industrial fabrication process.
4H-SiC allows the design of high voltage devices with a high integration level thanks to its 3.3 eV wide band gap, 2.5 MV.cm−1 high critical electric field and carriers mobility in the same order of magnitude as the one in sil- icon. Proposed diode can be implemented as free-wheeling diodes in an inverter leg config- uration.
After some technological considerations on design and fabrication, electrical characteriza-
tion will be presented. The static characteriza- tion first focuses on determining the forward current density, reverse voltage sustainability and temperature influence. Dynamic behavior is then investigated, in order to estimate turn- on and turn-off times in a buck-like configu- ration. Finally, surge current capabilities are estimated.
Past works on edge termination lead to re- liable process options [1]. So the focus was on the active area of the device, leading to a reduced number of technological steps.
1. State of the art of power diodes
Schottky and bipolar diodes have already been fabricated with a 3.3 kV voltage capa- bility [2] in ambient temperature as high as 300◦C. Since the Schottky diode on-state resis- tance is quite high, the junction area for a given forward current must be larger than bipolar diodes one.
Since diode devices compatible with a JFET process [3] have been investigated, reliable 3.3 kV JBS diodes have been demonstrated [4]
combining low threshold voltage of a Schot- tky diode and the low on-state resistance of the bipolar counterpart. Moreover, in reverse mode, such diodes present a leakage current 10 times lower than Schottky diodes one. Such a technology is now fully understood and per- fectly analyzed [5].
As a comparison, ST provides an example of state-of-the-art medium voltage (1200 V) commercial diodes [6]; the device is capable of 12 A forward current with short switching times. This paper focuses on bipolar-Schottky ratio for JBS diode and its influence on electri- cal performances.
2. Technological considerations
2.1. Process steps
First of all, the first epilayer was optimized with 1015cm-3 doping level and 40µm thick- ness [7] on 300CREE SiC material. For this op- timization, the edge termination and the passi- vation efficiencies were taken into account.
A 7-step process is needed to fabricate the proposed JBS diodes. It begins by a p+- implantation for the ohmic contact, followed by a p-implantation forming the edge termina- tion. In order to process a channel stopper, a n+ implantation is then performed around the edge termination. A first deposited oxide is opened where a thin and a thick metal creates the anode contact. Then comes a polymide passivation.
2.2. Devices specific features
In order to determine the influence of the bipolar part on JBS electrical characteristics, three different kinds of diodes were imple- mented on each elementary scribeline : two JBS diodes with 3 µm of p+ layer and 4 µm of n layer (JBS3-4) and one JBS diode with 3µm of p+layer and 8µm of n layer (JBS3-8).
The active area for both diodes is 3.9 mm2. A cross section of those devices is presented in figure 1. In the last field is a Schottky diode, without p+layer in its active area, but only on the edge, before the protection step process.
This scribeline plane leads to the fabrica- tion of more than 150 JBS3-4 diodes and 70 JBS3-8 diodes and Schottky diodes per wafer respectively. Such an important number of diodes fabricated following mature technolog- ical steps is compatible with industrial produc- tion related to a specific application but since the wafer still contains various geometries and test structures, it enables specific measure- ments in order to estimate where barriers re- mains.
2
Figure 1: JBS diodes with 3µm of p+layer and 4µm of n layer (left) and with 3µm of p+layer and 8µm of n layer (right)
A first run counts 3 wafers and measure- ments on the devices from these latter wafers are presented in next section.
3. Measurements, results and discussion
Measurements performed on those devices will lead to the complete knowledge of the be- havior of diodes in static and switching mode, temperature drift and stress. Moreover it will allow estimating design influence on elec- trical characteristics, specifically the bipolar- Schottky ratio on JBS.
3.1. High voltage capability
All diodes were first measured in forward bias mode in order to estimate the reliability of the front and rear ohmic contacts. Then the voltage capability was tested. The breakdown criterion is first set to 400 V because measure- ments are performed in air atmosphere. This ensures a rectifying contact without significant leakage in reverse bias, whereas it drives cur- rent in forward mode.
Table 1: Quality factor and Schottky part for diodes Kind of diode n (quality factor) Schottky area
Schottky 1.05 3.90 mm2
JBS3-8 1.06 1.98 mm2(51%)
JBS3-4 1.12 1.23 mm2(32%)
Based on these primary results, wafers were submitted to a higher voltage than theoretical limits but under vacuum atmosphere at room temperature [8].
As a first result, the influence of the p+ strips on breakdown voltage is analyzed: their presence or not for Schottky diodes and their spacing for JBS. Typical electrical characteris- tics are presented in figure 2.
Since avalanche is often not visible with this kind of diodes, the experimental breakdown criterion is set to 250 µA reverse leakage cur- rent. As it can be seen in figure 2, for a given voltage, there is one order of magnitude in leakage current between JBS3-4 devices and Schottky diodes.
Figure 2: Influence of bipolar part on leakage cur- rent (wafer level measurement in vacuum atmosphere, 250µA leakage current criterion, ambient temperature)
Overall for a higher bipolar part, diodes sus- tain higher voltage and present lower leakage current.
Thanks to the high number of available sam- ples, reliable statistics presented in figure 3, emerged from the measurements.
This figure shows that the majority of Schot- tky diodes (without p+ layer) or JBS3-8 (with a minor part of p+layer) are limited to 3 kV.
In contrast, since about half of the JBS3-4
Figure 3: Histogram of breakdown voltage of diodes (50µA leakage current criterion, vacuum atmosphere, ambient temperature)
devices with a thicker p+ layer sustain more than 3 kV, the importance of the bipolar on high voltage capability is confirmed.
3.2. Influence of temperature
Diodes were then diced and packaged in TO3 cases once all components static charac- teristics were extracted in forward and reverse modes for the whole working area. Compo- nents can now be submitted to higher current density and higher voltage in normal atmo- sphere and higher temperature.
Three characterizations were performed.
First a static characterization under warm pulsed air 25, 75, 125, 175 and 225◦C.
As presented earlier, all diodes have a more or less important Schottky part. Classical ef- fects of temperature on such diodes electrical characteristics in forward mode are a decrease in the threshold voltage and an increase in the on-state resistance; figure 4 presents this lat- ter effect, leading to a decrease of the forward current for a given voltage.
A typical effect of temperature in reverse mode is an increase leakage current. figure 5
Figure 4: Influence of temperature on different kinds of packaged diodes (JBS3-4, JBS3-8 and Schottky re- spectively) in terms of on-state resistance drop, forward current forVAK=4 V and voltage drop forIA=3 A
4
illustrates this effect, showing the drop of the current for a given voltage of−1000 V.
Figure 5: Influence of temperature on leakage current for different kinds of packaged diodes (JBS3-4, JBS3-8 and Schottky respectively) under 1 kV reverse voltage
At room temperature, threshold voltages are the same for all diodes, likewise at 225◦C.
Nonetheless, if JBS3-4 or JBS3-8 diodes have about the same on-state resistance, Schottky diodes on-state resistance is about 25% lower than JBS devices’ one at room temperature as well as at 225◦C.
3.3. Robustness and reliability
Secondly, the robustness of the diodes in conduction was tested with a 4 A forward pulsed current during 1 h, repeated 5 times.
The pulses are 100µs long and repeated every 200 ms. After each 1 h period, a static char- acterization was performed to estimate a pos- sible degradation. As shown in figure 6, there is only minor degradation. As a conclusion, it can be said that all diodes are able to drive a high forward current.
During and after the test, the threshold volt- age and the on-state resistance show the same experimental values as before. So a 5 h stress under high current is not harmful for diodes so far. The typical self-heating phenomenon in
Figure 6: Static characterization during and after robust- ness test, consisting of a 4 A forward conduction during 1 h, repeated five times, and relative variations of the current atVAK=4 V after the test
the Schottky is here avoided thanks to the use of pulsed current.
3.4. Summary: prefered geometry for im- proved static characteristics
Presented results on the high voltage capa- bility, the behavior under different tempera- tures or the robustness shows a trend for the in- fluence of bipolar/Schottky ratio on JBS static electrical characteristics. As a first conclusion it has been said that even if Schottky diodes have a lower on-state resistance (figure 4), if the bipolar part is important, the device sus- tain higher voltage with less leaks (figure 2).
This phenomenon is more true for a high tem- perature ambiance in reverse mode (figure 5).
Robustness of the diodes is not engaged, so for a high voltage reaching, the preferred de- sign will be the one with a larger bipolar part : the JBS3-4 diode, even if it is a little bit more resistive.
3.5. Dynamic characteristics and influence of geometry
Figure 7 presents a C-V characterization of the three types of diodes. Voltage varies from
0 down to−40 V in order to offer a correlation with switching mode results, with 20 mV am- plitude, at 10 kHz. Under−20 V, capacitances are the same for all types of diodes.
Figure 7: C-V characterization for junction area estima- tion
For low voltages (until−20 V), the diode ca- pacitance is higher if the bipolar part is larger.
The phenomenon is due to the small space- charge region around p+bulks. Indeed, the ra- tio written between JBS3-4 and Schottky area, and between JBS3-8 and Schottky, respec- tively 2.25 and 1.71 is the equivalent junction area including the 0.5µm depth of p+bulks.
But it must be noted that under a −20 V reverse-voltage applied to diodes, capaci- tances are the same for all types of diodes. It means that for higher voltage, the junction area is the same for JBS and Schottky diodes. So, the JBS traditional bipolar part is not activated under high reverse voltage.
This has a strong influence on switching performance : a smaller junction area means less charge, so the diode presents a lower switching time. This will be estimated with the circuit presented on figure 8, used for the last characterizations in switching mode.
In this conditions, for low voltage switch- ing, JBS3-4 will be faster. In the other hand, for high voltage switching, for which one
Figure 8: Schematic of the buck-like configuration used for diode switching behavior analysis, under 40 V and 1 A
diodes were designed, geometry and bipolar part width has no influence.
All diodes have mostly the same behavior when working in commutation. figure 9 shows the turn-off waveform for a JBS3-4 device, a JBS3-8 device and a Schottky diode. The dif- ferences between the waveforms are minor as summarized in table 2.
Figure 9: Turn-offswitching waveforms for a JBS3-4 device, JBS3-8 device and a Schottky diode, under 40V and 1A in the circuit in figure 8
The trr value is the time between the first sign change in the current and the time the cur- rent meets 0 A after its first maximum. This value and Qrr one are in the same order of magnitude as those of medium voltage com- mercial diodes [6]. The JBS3-4 Qrr is the lower, thanks to its low junction area that en- ables a fast switch at high dI/dt.
6
Table 2: Comparison of main electrical characteristics for different types of diodes
Schottky JBS3-8 JBS3-4
VAK-max 55 V 53 V 52 V
IA-max 4 A 3.7 A 3.4 A
Qrr 630 nC 550 nC 500 nC trr 730 ns 585 ns 612 ns dI/dt 136 A/µs 167 A/µs 162 A/µs
3.6. A destructive measurement : surge curent measurement
Barcelona’s microelectronics center (CNM- IMB) enables high current characterization thanks to a home-made tester [9], applying the classical half-sine wave on the DUT. Since this kind of measurement is destructive for the de- vice, surge current capabilities was estimated at the end. Figure 10 presents the results of the measurement at the breakdown point. The breakdown current is summarized in table 3.
Figure 10:I−Vmeasurement at the breakdown current for the three kind of diodes
It can be seen on figure 10 that for a volt- age higher than 20 V, the current first decreases before breaking down. A high on-state resis- tance can lead to a self-heating of the diode and a self-increase of the on-state resistance.
This maybe results from the device as well as
Table 3: Summary of the breakdown current for different types of diodes
Kind of diode IBR 0.8×IBR Schottky 18.9 A 15.1 A JBS3-8 17.4 A 13.9 A JBS3-4 17.6 A 14.1 A
the package of the diode [10]. Traditionally, JBS diodes have a higher surge current capa- bility thanks the bipolar injection. But what is also induced by this figure is that the p+layer is not activated, and thus cannot inject current through the diode [11]. In this case, Schot- tky diode has a higher current capability and a lower on-state resistance even when working at high current.
Conclusion
The existing 3.3 kV diode market is not widely covered with dedicated devices. De- vices presented in this paper can be considered as a reliable answer for inverter applications, thanks to their high voltage capability.
Moreover, high performance JBS diodes al- low a high forward current density and a fast switching time.
In facts, JBS3-4 will be preferred for its high voltage capability (more than 3.5 kV) with low leaks (less than 50 µA). Thanks to the low influence of design on switching delays, this diode will be fast enough for many high volt- age applications.
Acknowledgement
Authors want to thanks DGA, the French Weapon Delegation for the financial support of this work, the ISL, French-German research institute for the measurement facilities, and the CNM-IMB for surge current measurements.
References
[1] P. Brosselard, D. Planson, S. Scharnholz, C. Ray- naud, V. Zorngiebel, M. Lazar, J.-P. Chante, E. Spahn, Solid-State Electronics 50 (2006) 1183–
1188.
[2] B. Vergne, G. P¢ques, C. Maurer, S. Scharnholz, P. Brosselard, D. Planson, in: EPF2008.
[3] P. Brosselard, D. Tournier, M. Vellvehi, J. Montserrat, P. Godignon, J. Millan, Mate- rials Science Forum 556-557 (2007) 1003–1006.
[4] P. Brosselard, V. Banu, N. Camara, A. Prez- Tomas, Material Science and Engineering B 165 (2009) 15–17.
[5] R. Radhakrishnan, J. Zhao, Solid-State Electron- ics 63 (2011) 167–176.
[6] S. Microelectronics, STTH1212, ultrafast recov- ery 1200 V diode, 2006.
[7] S. Sze, Physics of Semiconductor Devices, New York: Wiley, 1981.
[8] B. Vergne, G. P¢ques, C. Maurer, S. Scharnholz, P. Brosselard, D. Planson, in: EPF2012.
[9] V. Banu, P. Godignon, X. Jord´a, X. Perpinya, J. Mill´an, Materials Science Forum 740-742 (2013) 873–876.
[10] J. Hilsenbeck, M. Treu, R. Rupp, K. Rschen- schmidt, R. Kern, M. Holz, Materials Science Fo- rum 645-648 (2010) 673–676.
[11] M. Treu, R. Rupp, C. Tai, P. Blaschitz, J. Hilsen- beck, H. Brunner, D. Peters, R. Elpet, T. Reimann, Materials Science Forum 527-529 (2006) 1155–
1158.
8
![[PDF] Programmation du système Unix en Objective Caml | Cours informatique](data:image/gif;base64,R0lGODlhAQABAIAAAP///wAAACH5BAEAAAAALAAAAAABAAEAAAICRAEAOw==)